What AI can an RTX 3060 12GB run?

Quick verdict

PC with RTX 3060 12GB (12 GB) and 32 GB RAM can run 14 of the 31 models tracked here at a comfortable speed, and another 9 slowly. The largest chat or coding model that fits is Phi-4 14B, which needs about 10.1 GB and generates roughly 18 words a second. The quickest is Gemma 4 E2B at about 81 words a second, which feels faster than you can read.

What limits this machine

Video memory (VRAM) is the limit: 12 GB on this card. A model must fit in VRAM to run fast. Anything larger spills into system RAM (32 GB in this setup), which works but is slow, because system RAM moves data at roughly 70 GB/s while the card moves 360 GB/s.

What it is good for

Chat and coding

Best chat pick: Gemma 4 12B at about 20 words a second. Best coding pick: Qwen2.5 Coder 7B at about 35.

Images

• Stable Diffusion XL: runs well, about about 18 s per image
• FLUX.2 klein 4B: runs well, about about 34 s per image

Video

Video models need more memory than this machine has.

Voice and music

• Whisper large-v3 turbo: runs well, about about 1 min per 10 minutes of audio
• Kokoro TTS: runs well, about about 11 s per minute of speech
• Qwen3 TTS 1.7B: runs well, about about 52 s per minute of speech
• ACE-Step 1.5: slow, about about 5 min per 1 minute song

Fastest models here

• Gemma 4 E2B: about 81 words a second, faster than you can read
• Gemma 4 E4B: about 51 words a second, faster than you can read
• Qwen 3.5 4B: about 49 words a second, faster than you can read
• Qwen2.5 Coder 7B: about 35 words a second, smooth
• DeepSeek R1 8B: about 32 words a second, smooth

Memory headroom

Spare memory after loading the model decides how long a conversation or document can be before things slow down. With about 31 GB usable here:

• Phi-4 14B needs 10.1 GB, leaving 20.5 GB spare, plenty for long documents.
• Gemma 4 12B needs 8.9 GB, leaving 21.7 GB spare, plenty for long documents.
• Gemma 4 E4B needs 7.4 GB, leaving 23.2 GB spare, plenty for long documents.
• Qwen 3.5 9B needs 7.4 GB, leaving 23.2 GB spare, plenty for long documents.
• Qwen3 VL 8B needs 6.9 GB, leaving 23.7 GB spare, plenty for long documents.

What does not fit, and by how much

• Llama 3.3 70B needs about 45.6 GB and this machine tops out near 31 GB, so it is about 15 GB short.
• gpt-oss 120B needs about 68.8 GB and this machine tops out near 31 GB, so it is about 38 GB short.
• Qwen 3.5 122B A10B needs about 85.5 GB and this machine tops out near 31 GB, so it is about 55 GB short.
• DeepSeek V3 671B needs about 424.7 GB and this machine tops out near 31 GB, so it is about 394 GB short.

How it compares

• PC with RTX 4060, RTX 3070, RTX 5060, RTX 4060 Laptop or RTX 4070 Laptop and 32 GB RAM: 5 fewer models; drops Gemma 4 12B, Phi-4 14B, Stable Diffusion XL and more.
• PC with RTX 3080 10GB (10 GB) and 32 GB RAM: 2 fewer models; drops Phi-4 14B, FLUX.2 klein 4B.
• PC with RTX 4070/4070 Super or RTX 5070 and 32 GB RAM: the same number of models.
• PC with RTX 4060 Ti 16GB (16 GB) and 32 GB RAM: 2 more models; adds gpt-oss 20B, ACE-Step 1.5.

Is an upgrade worth it?

The biggest single jump is Ryzen AI Max+ 395 (Strix Halo), 128 GB, which would unlock 13 more models, including gpt-oss 20B, Gemma 4 26B A4B, Qwen 3.5 35B A3B.

Questions